Image Reconstruction Algorithms For Limited-angle Computed Tomography

X-ray computed tomography?CT?is an effective imaging technique that can obtain the internal information of the scanned object.In recent decades,CT has been widely used in many fields,such as industrial testing,clinical medicine and security testing.As one of the key steps in X-ray CT,image reconstruction uses the projection data obtained by the detector to reconstruct the tomographic image of the scanned object.When the projection data collected by the detector are complete?for example,in fan beam CT,if scanning rotation angle range is greater than or equal to 180^o plus fan angle,the projection data are complete?,the Filtered Back Projection?FBP?algorithm,the most widely used algorithm in commercial CT,can reconstruct high quality images.However,in some practical applications?such as the imaging of breast CT?C-arm CT?dental CT and CT imaging of pipeline in service?,due to the limitations like the structure of scanned object,the scanning environment and the X-ray radiation dose,the projection data can only be obtained from a limited scanning rotation angle range,hence those projection data are incomplete.When dealing with the above problems occurred in image reconstruction for limited-angle CT,image reconstructed by the FBP algorithm will lead to artifacts near the edges.To improve the reconstructed image quality of limited-angle CT is a hot topic in CT image reconstruction with high academic significance and important application value.This dissertation focuses on solving the problem of image reconstruction for limited-angle CT.First,we analyze the existing limited-angle CT image reconstruction algorithms.Then,we systematically study the drawbacks of these algorithms.Final,we improve these limited-angle CT image reconstruction algorithms with some intelligence techniques to reduce the limited-angle artifacts,enhance the structural protection ability and improve the robustness to noise.The main work of this paper is as follows:1.Under the limited-angle CT scanning mode,due to insufficient projection data obtained,most image reconstruction algorithms will lose information about the edge structures and the details of the reconstructed image.In this situation,although the results of the SART method show severe noise and limited-angle artifacts,the information of details and structures still exist.The results of the image reconstruction algorithm based on the wavelet frame L₀ quasi norm can better suppress noise and reduce limited-angle artifacts,whereas the information of details and structures are over-smoothing.In addition,Guided image filtering?GIF?,a smoothing operator with boundary protection,can make the filtered image has similarity with guidance image by transferring the important features of the guidance image into the input image.In order to reconstruct the high quality image in the limited-angle CT scanning mode,a limited-angle CT image reconstruction algorithm based on GIF and wavelet framework L₀ quasi norm is proposed.In each iteration of the proposed algorithm,the reconstructed result gained by using the image reconstruction algorithm based on the wavelet frame L₀ quasi norm is used as the guidance image,GIF will transfer the important features the guidance image contained into the reconstructed result of SART method.Some simulated projection data experiments and real data projection data experiment are conducted to test the feasibility and effectiveness of the proposed algorithm.The qualitative and quantitative indexes indicate that the proposed algorithm is superior to two other iterative reconstruction algorithms in artifacts reduction,noise suppression and structure preservation.2.As a low-end CT system,parallel translational CT?PTCT?is in urgent demand in developing countries.Under some circumstances,in order to reduce the scan time,decrease the X-ray radiation dose or scan long object,we use the limited-angle PTCT scanning mode to scan an object within a limited rotation angle range.However,this scanning mode introduced some limited-angle artifacts that seriously degraded the imaging quality.To reconstruct a high-quality image in the limited-angle PTCT scanning mode,we propose a limited-angle PTCT image reconstruction algorithm based on deep learning.First,we use the SART method on the limited-angle PTCT projection data.Then we import the SART result to a well-trained deep convolutional neural networks?CNN?which can reduce the artifacts and preserve the structures to obtain a better reconstructed result.Some simulated projection data experiments are implemented to test the feasibility and effectiveness of the proposed algorithm.By comparing with three state-of-the-art methods,the proposed algorithm can effectively preserve the image structures,suppress the noise and reduce the limited-angle artifacts.3.Regularized image reconstruction algorithm can better deal with the limited-angle CT image reconstruction problem,but these algorithms are difficult to find an appropriate regularization term and adjust the regularization parameters.In addition,when the range of the scanning rotation angle is small,the image reconstructed by the regularized image reconstruction algorithm will not be satisfactory.To solve these problems,we propose an alternating direction method of multipliers?ADMM?-based deep reconstruction?deep learning based image reconstruction?algorithm for limited-angle CT.First,we use the ADMM algorithm framework to solve a regularized image reconstruction mode.Then,we utilize a deep CNN to replace a part of the ADMM to reduce limited-angle artifacts and avoid the choice of the regularization term and the adjustment of the regularization parameters.The feasibility and superiority of the proposed algorithms are verified by some simulated projection data experiments.The experimental results indicate that our algorithm has a better performance than two state-of-the-art algorithms at structure preservation and artifact reduction.